The most dangerous move in chess is castling. Sort of.

We keep an engine’s verdict on every game our players import. For each game, the single move that cost the most. Group those worst moves by their notation and one string dominates: O-O, castling short, top of the list for 56 of 223 players. Read naively, castling is the most dangerous move in club chess.

We nearly published that sentence. It is true, and it is wrong, and the gap between those two things is the whole reason this column exists.

Notation is a counting trick

O-O is the only move in algebraic notation with a name of its own. Every knight move is split across dozens of spellings. Nf3, Nxe5, Nbd2. and every queen move likewise. Castling wins the count because it is the only candidate that is not divided against itself. It is the electoral math of chess notation: one unified party against a hundred splinter groups.

Group blunders by the piece that moved instead and the table turns over completely. The queen leads at 19.8% of a typical player’s blunders, the knight just behind, and castling collapses to 1.4%. The smallest category there is. It is the worst piece for exactly zero of 223 players.

What we changed because of it

Both figures now live in the product, each labelled as what it is: the notation count as a curiosity about how chess is written, the per-piece count as the thing to train against. And our publishing pipeline gained a rule we recommend to anyone who computes over games: before you print ‘the most common X’, check whether your categories are the same size. Ours were not, and the smallest category was wearing a crown.